Deep learning-based T-wave form classification system with strong generalization ability

Abstract

The invention provides a deep learning-based T-wave form classification system with a strong generalization ability. The system comprises a data preprocessing module which is configured to obtain electrocardiosignal data and carry out the preprocessing of the obtained data, a fragment extraction module which is configured to extract the wave peak position of R waves and the end point position of T waves to obtain a signal fragment containing the R waves and the T waves, a prediction probability vector extraction module which is configured to convert the signal fragment into a time domain image and a time-frequency domain image, input the time domain image into a first convolutional neural network to obtain a time domain prediction probability vector, and input the time-frequency domain image into a second convolutional neural network to obtain a time-frequency domain prediction probability vector, and a classification module which is configured to perform weighted fusion on the prediction probability vectors of the time domain and the time-frequency domain, and obtain a T-wave form classification result according to the newly obtained vectors. According to the system, the accuracy of T-wave form classification is greatly improved for dynamic wearable electrocardiogram monitoring or strong-noise electrocardiogram signals.

Description

Technical field [0001] The present disclosure relates to the field of electrocardiographic signal classification, and in particular, to a T wave form classification system based on deep learning. Background technique [0002] The statement of this section is merely the background technology related to the present disclosure, and it is not necessarily constituting the prior art. [0003] Cardiovascular disease has become a primary reason for human death. As the wearable technique continues to advance, the long-term continuous monitoring of the electrocardiographic signal is possible. The dynamic electrocardiogram is a method of continuously record and analyzing the changes in the core map of the human body. Generally, the electrocardiographic signals up to 100,000 times can be continuously recorded within 24 hours, which can improve the non-sustained arrhythmia, especially The detection rate of arrhythmia and short myocardial ischemia, expanded the scope of the clinical applicatio...

AI Technical Summary

Problems solved by technology

[0007] However, the above-mentioned algorithm based on artificial features has greatly affected the robustness of the algorithm, and the algorithm recognition accuracy has also encountered a bottleneck.

Images